A Two-Stage Dynamic Parameters Identification Approach Based on Kalman Filter for Haptic Rendering in Telerobotics

In teleoperated physical Human-Robot Interaction (pHRI), accurate contact force feedback is essential for operators to experience intuitive and responsive interaction. However, the presence of end-effectors on the manipulator often interferes with the force feedback, misrepresenting the real contact forces. To render a precise force to operators, it is critical to eliminate the non-contact forces arising from the end-effector tool's inertia. To achieve this, the fast and accurate dynamic parameters identification for the end-effectors is essential. Currently, existing dynamic identification algorithms process all collected data in one stage, in which the sensor noises for static and dynamic parameters are coupled, resulting in poor performance. In this letter, we propose a two-stage method for end-effector’s inertial parameters identification. Our proposed method breaks the identification process into two steps: the static and the dynamic processing stages, which can make full use of data features and significantly improve the accuracy of mass and Center-of-Mass (COM) identification. Additionally, we introduce the Kalman Filter (KF) into the linear fitting to improve the identification speed. Comparative and ablation experiments are designed to validate the effectiveness of the proposed method. The results demonstrate that our method outperforms the RTLS method in reducing identification errors by 96% for mass and 81.2%–87.6% for COM. As for the haptic rendering, the errors are also reduced by 36.3% in direction of interaction and 99.5% in direction of gravity after the elimination of the end-effector’s inertia.